An evaluation method based on neural network is put forward for the purpose of predicting subjective perceptions of automobile comfort seat. The inputs includs 12 seat interface pressure measures and two anthropometrics. The output is an overall comfort index derived from a survey. The SVR is developed and validated by using data collected from 12 subjects，and the subjects is evaluated by five different driving seats. The prediction results reach 0.0030 in mean square error and 0.749 in linear correlation coefficient， it gives better results than that with artificial neural network. The resulting knowledge can reduce the cost and time associated with the current automobile seat comfort development process.